The WWP uses <mw type="catch"> to encode catchwords. We do not nest any encoding within the <mw> element (except <sic> as noted below).
The WWP encodes any discrepancies between the catchword and its corresponding word at the top of the next page, using <sic>, without a corr= attribute. However, if the catchword reprints only part of the corresponding word, we do not use <sic>, but ignore the difference, since this is a fairly common practice and probably not to be considered an error. We also would not encode with <sic> any differences in capitalization or rendition between the catchword and the corresponding word. See example 1.
If the catchword prints an entire word, of which the text (erroneously) prints only part, then a <sic> element *in the text* should indicate the error and its correction (using <sic> with a corr= attribute). See example 2.
The WWP encodes any hyphens within catchwords as ordinary hyphens (using the character “-”). We do not encode them as soft hyphens, since they are not intended to disappear if line breaks are removed. Quotation marks and other delimiters are encoded as #PCDATA rather than by using the rend= attribute, since there are no circumstances under which we would want to vary their presentation.
Example 1. Error appearing
in catchword (discrepancy between catchword and text)
last word on preceding page: “un-”; catchword: “pining”; first word on next page: "repining"
Example 2. Error in text
last word on preceding page: “un-”; catchword: “repining”; first word on next page: "pining"
Example 3. Error in both catchword and text
last word on preceding page: “un-”; catchword: “repinning”; first word on next page: "pining"